Automated marketing reporting: track what pays off

On Monday morning, someone opens five tabs: Google Analytics, Search Console, the email marketing tool, the ad platform and the CRM. They copy figures into a spreadsheet, format them, and the morning is gone. The report that comes out says how many visits, clicks and opens there were. It rarely says what management wants to know: what paid off, and what to change.

This guide builds on our guide to marketing automation for SMEs, which places reporting among the other marketing tasks. It is written for teams of 20 to 500 people who already have their tools and no longer have time to copy out their figures.

Why automate your marketing reporting?

Because a report copied out by hand has three flaws, and none of them comes down to the talent of the person doing it. Automating means writing the rule once: which figures, taken from which tool, with which definition, sent when and to whom.

Which figures should you track to know what pays off?

A useful report reads from top to bottom, from the figure that talks about money to the one that explains it.

1. What pays off: contact enquiries, quotes, signed deals and their value. These figures live in the CRM, not in the analytics tool. That is why the CRM is the centrepiece of any reporting set-up: we have used Twenty in production since early 2025, and it is what shows where each deal comes from.
2. What it costs: ad spend and time spent, set against the number of new customers. A channel that brings in many visits and no sales remains an expensive channel, even when each click costs nothing.
3. What explains it: useful visits (those that reach an offer page or a form), clicks from Google and the queries that bring them, clicks from emails. For an email campaign, judge by clicks, replies and unsubscribes: as our guide to running an email campaign explains, the open rate is no longer a reliable indicator.
4. What reassures without telling you anything: the number of subscribers, impressions or page views. These figures can appear at the bottom of the report, never at the top.

Each figure has one written definition, and only one. Is a contact enquiry a form submitted, a call, an email received? As long as the answer changes depending on who is counting, two reports cannot be compared.

Where do the figures come from, and how do you connect them?

The figures in a marketing report come from four or five tools that do not talk to each other: the audience measurement tool (Google Analytics or Matomo), Google’s Search Console for organic search, the email marketing tool, the ad platform and the CRM. Connecting them takes two things.

Tagged campaign links. Every link you put out, in an email, a post or an ad, carries parameters that say where the visit came from. Google Analytics help is explicit: “When you add parameters to a URL, you should always use utm_source, utm_medium, and utm_campaign” (Google Analytics help on campaign URLs), that is, the source, the medium and the campaign. Without them, part of the visits from your newsletter arrive with no known origin, and what they earned is lost.

A source recorded on every contact. The form that creates a contact in the CRM copies these parameters onto its record. The day the deal is signed, you know which channel it came from, and the value of the sale can be sent back to the ad platform so that bidding follows what sells.

The rest is plumbing. An automation tool such as n8n, which we run on our own server alongside Twenty and Matomo, can query each tool on Monday morning, assemble the figures and send the report. Nothing is copied, so nothing is copied wrongly.

What can AI do in reporting?

It is useful in one place only: after the figures. A language model given this week’s report and the previous week’s can write, in three sentences, what has moved. For example: “Enquiries coming from Google have halved, all of them on a single page, which has lost its place in the results since Tuesday.” Management reads that sentence; it does not read a forty-line table. For this job, an inexpensive model is usually enough: commenting on a table does not call for the most powerful model on the market.

Three rules prevent unpleasant surprises.

• The figures come from the tools, never from the model. You give it figures that are already calculated, and it comments on them. A model that adds up, or “estimates” a missing figure, can get it wrong without saying so.
• Each comment cites the figure it comments on, so that it can be checked at a glance.
• A person decides. AI can suggest a lead, such as reviewing a page that is slipping; it does not move a budget.

This is how our SEO engine runs:

The counting calls on no model; the judgement stays with a person.

Which data pitfalls should you know about before you start?

How long data is kept. Tools do not keep everything, and not for ever. In a standard Google Analytics 4 property, user-level and event data are kept for two or fourteen months, depending on the setting chosen, and up to fifty months for properties on Analytics 360. This setting “does not affect standard aggregated reports”: it only affects explorations and funnel reports (Google Analytics help on data retention).

Visitor consent. An audience measurement tool generally places trackers, and the law requires the visitor’s consent, with exceptions. The CNIL, France’s data protection authority, exempts the trackers used for “a purpose strictly limited to the sole measurement of the site’s or application’s audience”, which serve to produce “anonymous statistical data only” and are not cross-referenced with other processing; their lifespan remains limited (it gives thirteen months as an example) and the information collected is “kept for a maximum of twenty-five months” (CNIL, audience measurement). These conditions depend on how the tool is configured, not on its name: it is the configuration you must check before removing a consent banner.

Figures that do not match. Analytics, the ad platform and the CRM do not count the same thing: a conversion in one, a click in another, a deal in the third. Do not try to make them coincide. For each figure, choose the tool that has the final say, and write it down. For sales, it is the CRM.

Where to start?

Three steps, in this order:

1. 01Write the list of figures.

   Keep five figures at most, with for each its definition, the tool that has the final say and the person who reads it. This choice takes a short meeting, and the list decides everything else.

2. 02Connect the tools.

   Tag your campaign links, record the source on every contact, and give the automation read access to the tools.

3. 03Automate the sending.

   Have the report go out every Monday, and only then add the AI’s comment.

If you open new markets, track each one separately from the start, with the threshold that will say whether it deserves to be kept: a language that grows can hide one that is declining. This is the method in our article on international SEO and the choice of markets.

In our digital marketing in one block offer, this report is delivered every month: what each channel earned in contacts, deals and sales, with the sales forecast.

Frequently asked questions

Sources

1. Google Analytics help on campaign URLs (new tab)support.google.com
2. Google Analytics help on data retention (new tab)support.google.com
3. Search Console help on bulk export (new tab)support.google.com
4. CNIL, audience measurement (new tab)cnil.fr